Dissertations, Theses, and Capstone Projects CUNY Graduate Center
5-2018
Perceived Knowledge and Defense of Political Attitudes Perceived Knowledge and Defense of Political Attitudes
Matthew H. Goldberg
The Graduate Center, City University of New York
How does access to this work benefit you? Let us know!
More information about this work at: https://academicworks.cuny.edu/gc_etds/2662 Discover additional works at: https://academicworks.cuny.edu
This work is made publicly available by the City University of New York (CUNY).
Contact: [email protected]
PERCEIVED KNOWLEDGE AND DEFENSE OF POLITICAL ATTITUDES by
MATTHEW H. GOLDBERG
A dissertation submitted to the Graduate Faculty in Psychology in partial fulfillment of the requirements for the degree of Doctor of Philosophy, The City University of New York
2018
©2018 MATTHEW H. GOLDBERG All Rights Reserved
Perceived Knowledge and Defense of Political Attitudes by
Matthew H. Goldberg
This manuscript has been read and accepted for the Graduate Faculty in Psychology in satisfaction of the dissertation requirement for the degree of
Doctor of Philosophy.
Date Cheryl L. Carmichael, Ph.D.
Chair of Examining Committee
Date Richard Bodnar, Ph.D.
Executive Officer
Supervisory Committee:
Curtis D. Hardin, Ph.D.
Steven G. Young, Ph.D.
Kimberly Rios, Ph.D.
THE CITY UNIVERSITY OF NEW YORK
Abstract
PERCEIVED KNOWLEDGE AND DEFENSE OF POLITICAL ATTITUDES by
MATTHEW H. GOLDBERG Advisors: Cheryl Carmichael & Curtis Hardin
Three experiments tested if perceived knowledge about a political issue predicted people’s willingness to engage with relatively weaker versus stronger belief-threatening information on that issue. Study 1 assessed people’s perceived knowledge on four political issues and, for each issue, manipulated whether participants chose between arguing against a weak versus moderate argument or a moderate versus strong argument. Only one issue provided some support for the primary hypothesis. When people believed they were not knowledgeable about the carbon tax, giving them a relatively stronger set of arguments to choose from (moderate versus strong) further increased their preferences for the weaker argument compared to an easier set of choices (weak versus moderate). When people believed they were high in knowledge of the carbon tax, the effect disappeared, demonstrating that they were more likely to engage with strong belief- inconsistent information. Study 2 tested whether perceived knowledge predicted people’s choices of whether to counter-argue versus ignore weak versus strong belief-inconsistent information. Perceived knowledge predicted preferences for counter-arguing both weak and strong messages, but the relationship was slightly stronger for strong messages. There was also a strong effect of condition such that people became significantly more likely to ignore the
information when the information was strong than when it was weak. This effect was primarily driven by people’s expectations that they could not effectively counter-argue strong belief-
inconsistent information. In Study 3, people chose between two debate partners that were either similar in knowledge versus substantially lower in knowledge or between partners that were similar in knowledge versus substantially higher in knowledge. When people believed they knew little about the debate topic (i.e. the death penalty), their choices were unaffected by the available options; when people believed they knew a lot about the topic, they consistently chose the more knowledgeable debate partner. These findings suggest that people’s perceived knowledge and their beliefs about their ability to effectively counter-argue belief-inconsistent information play an important role in how they decide to defend their beliefs.
Acknowledgements
My journey to my doctorate started long before graduate admissions. As an
undergraduate at Hofstra University, I took a social psychology course with Sarah Novak and quickly learned that this was what I wanted to spend my life studying. There is no one more responsible for my admission to a doctoral program than Sarah. It took endless drafts of my personal statements, getting me the right research experience, and long talks about how much of my “true” self to reveal in the admissions process and graduate life. It turns out that the grit I developed through years of wrestling and mixed martial arts translated to a successful graduate career, where grit was needed most. It was Sarah who celebrated this aspect of me and made sure that I had the opportunities to prove myself in graduate school. Because of Sarah I am writing this, five years later, having earned my doctorate. Thank you so much for believing in me.
There was no one more influential in my graduate career than Cheryl Carmichael. You have gone above and beyond your role as an advisor to teach me what it means to be an excellent scholar, teacher, and mentor. Throughout all of the graduate school growing pains, you brought warmth and kindness—ensuring that I would accomplish every goal I set out to achieve. I will be forever grateful for all that you have done for me and it is no exaggeration when I say I could not have done this without you.
Thank you to Curtis Hardin for shaping my philosophy of science and inspiring so many of my ideas. You have always affirmed that my work was important and worth doing. It is because of you that I learned to think theoretically, giving me the ability to take seemingly disparate findings and integrate them into a unified framework. Your wisdom will always stick with me that the people who are successful in this field are those who do the work because they love doing it.
Thank you to Steven Young and Kimberly Rios for becoming my informal, but no less important, advisors. I am extraordinarily thankful to Steve for giving me countless opportunities to progress my career and for being one of the kindest people I know. Kim, thank you for welcoming me into your world of incredibly interesting and important research. It was through you that I first learned about the peer-review process and what it takes to be an excellent scholar in this field. Thank you so much, Steve and Kim, for being my mentors and for going out of your way to support me. I am so thankful to have met you.
Thank you to my brilliant colleagues for sharing this journey with me. I have learned so much from you and I hope I can provide you with the same support as you continue along this journey. Thank you to our faculty for creating an excellent environment for me to learn and grow. I am truly proud to be a product of the Basic and Applied Social Psychology program.
Special thanks to Sarit Golub for treating me like a respected colleague starting from day one.
Every time a major problem was on the horizon, it was you who went out of your way to fix it before it could ever affect me. It is because of you that my graduate career has gone so smoothly and for that I will be forever grateful.
Thank you to my brother, Michael Goldberg, and to those whom I consider my brothers, Jonathan Wagner and Salvatore David, for being part of this relentless pursuit of self-
improvement. Living life by your side has created my very definition of what it means to be great.
Thank you to my girlfriend, Alexis Robbins, for being an unending source of love and support. Your relentless pursuit of your own dreams constantly inspires me to keep chasing mine. Even when the hills were steep and the rejection letters were frequent, you never showed even an ounce of defeat. Because of you, I keep going.
I couldn’t be more grateful for my family for always believing in me. Thank you to my mother, father, and Avi for supporting my every move and for displaying character and work ethic I aspire to match. Thank you to my grandmothers, Bella and Florence, for keeping my life simple. You constantly remind me that all I really need is your love, support, and, of course, too much food.
As one could easily tell from these acknowledgements, I am unbelievably fortunate to have had all of these amazing people shape my life. As I grow, I hope to inspire each of you in all of the unique ways you have inspired me.
Table of Contents
Abstract………...iv
Acknowledgements………...vi
Table of Contents………...ix
List of Tables...………...x
List of Figures...………...xi
Chapter 1: Introduction...………...1
Chapter 2: Study 1………...8
Chapter 3: Study 2………...16
Chapter 4: Study 3………...23
Chapter 5: General Discussion………...29
Appendix A………...38
Appendix B………...43
Tables………...45
Figures………...47
References………...51
List of Tables
Table 1a. Study 1 Argument Strength Pretest Ratings: Weak vs. Moderate...43 Table 1b. Study 1 Argument Strength Pretest Ratings: Moderate vs. Strong...44
List of Figures
Figure 1. Study 1 Condition × Perceived Knowledge interaction predicting choice of carbon tax arguments...45 Figure 2. Study 2 Condition × Perceived Knowledge interaction predicting task choice...46 Figure 3. Mediation model testing the indirect effect of condition on choice of defense strategy through expectations of effective counter-arguing...47 Figure 4. Study 3 Condition × Perceived Knowledge interaction predicting partner choice...48
Chapter 1: Introduction
Voter ignorance is a widely recognized phenomenon in American politics (Arnold, 2012;
Althaus, 1998). In spite of their ignorance, people still find ways to defend their political beliefs from dissent. Although people can protect their beliefs by expending more effort criticizing belief-inconsistent information (Edwards & Smith, 1996; Ditto & Lopez, 1992) or avoiding it altogether (Eagly & Chaiken, 1993), successfully arguing against strong information provides a greater epistemic benefit than arguing against weak information because it increases attitude certainty (Tormala & Petty, 2002). The proposed research is designed to show that, as people’s self-perceived knowledge on an issue increases, their tendency to choose to approach strong belief-inconsistent information will increase. Additionally, I propose that people will approach the strongest belief-inconsistent information that they believe they can refute. That is, when people believe they are highly knowledgeable, they will choose to argue against strong rather than weak belief-inconsistent information because refuting strong information proves the superiority of one’s beliefs more than refuting weak information.
For example, when people believe they are high in political knowledge they should expect a high probability of successfully counter-arguing criticisms of their political views. Such people should be more inclined to approach strong belief-inconsistent information because counter-arguing strong information increases attitude certainty whereas counter-arguing weak information does not (Tormala & Petty, 2002; also see Tormala & Petty, 2004). People who believe they are low in political knowledge recognize their limited ways of defending their beliefs and should therefore only approach belief-inconsistent information when it is weak—
when they know that they can refute it.
Learning about how people defend their political attitudes is crucial to understanding how to effectively relay information to the uninformed electorate. However, the influence of
perceived knowledge on information search remains unclear. Thus, it is important to learn more about people’s willingness to engage with information that challenges their political position.
Cognitive Energetics Theory (CET; Kruglanski, Bélanger, Chen, Köpetz, Pierro, & Mannetti, 2012) provides a guiding theoretical framework for thinking about the kinds of information with which people seek to engage.
Cognitive Energetics Theory
CET uses the early terminology of Lewin (1936) to describe how motivated cognition predicts goal pursuit. That is, it assumes that goal attainment is guided by a driving force and a restraining force. A driving force represents the maximal amount of energy that a person is willing to invest in attaining a cognitive goal whereas a restraining force limits this energy.
Primary driving forces include the importance of attaining the cognitive goal and the amount of available mental resources. The most prominent example of this is at the core of the Elaboration Likelihood Model (ELM, Petty & Cacioppo, 1986) in which students read arguments in favor of imposing a comprehensive exam for all students at their university or at a distant university. For students in the condition that imposes the exam at their own university, it raises the driving force (i.e. personal importance or vested interest) and thus leads the students to deeply process the message content because the new exam will directly affect them. For students in the condition that imposes the exam at a distant university, there is little vested interest in the outcome, and therefore they will rely on shortcuts in evaluating the information (e.g. source expertise) rather than the argument content. This demonstrates that, when a task is of high importance, people will invest more energy in the task.
The amount of available mental resources is also a critical component of the driving force. For example, Kruglanski and Thompson (1999, Study 3) manipulated whether participants were to evaluate information under high or low cognitive load. The researchers then manipulated whether or not the information came from an expert source (expert vs. inexpert) and the length of the source background information (short vs. long). When under high cognitive load, participants relied more on the brief, but less informative, information to form their judgments whereas participants under low cognitive load relied more on the lengthy, but more informative,
information. Inducing high cognitive load reduced participants’ driving force towards attaining the goal of accurately judging the relevant information whereas people with the available mental resources (i.e. low cognitive load) were more willing to invest more energy in the task.
On the other hand, a restraining force inhibits goal pursuit via processes such as task difficulty or competing goals. For example, political partisans might be highly motivated (i.e.
high driving force) to defend their favorite candidate’s tax policy but will be limited in their ability to do so if they face strong counter-arguments from opposing partisans (i.e. high restraining force). Shah and Kruglanski (2002) activated a high driving force in students to perform well on an anagram task by telling them it measured “verbal recall,” which the experimenters described as a valuable skill for writing. The researchers demonstrate that, compared to priming task-related goals (i.e. performing well on the anagram task), priming alternative goals (i.e. increased restraining force) led to worse performance on anagram tasks.
Activating an alternative goal increased the restraining force on performing well on the anagram task.
Although CET states that the driving and restraining forces are separable, it also states that they are functionally equivalent. That is, for example, raising the restraining force is
analogous to reducing the driving force. CET states that when people are have a high driving force, they seek to choose the most effective way of attaining their cognitive goal even if that means investing more energy. For example, Pierro, De Grada, Mannetti, Livi, and Kruglanski (2004) demonstrate that people rely on cognitive shortcuts (e.g. consensus information) when they have a low driving force but they rely more on argument content when they have a high driving force. Relying primarily on consensus information is fallacious because it does not reveal anything about the arguments themselves. Put simply, people rely on this more efficient, but less valid, technique for understanding information when they are unmotivated to deeply process it (also see Petty & Cacioppo, 1986). On the other hand, when people are highly motivated (i.e.
high driving force), they will use more laborious ways of processing information (e.g. assessing message arguments) to the extent that is serves an appropriately important cognitive goal.
Because the primary goal of the current project is to investigate how perceived knowledge predicts people’s willingness to engage with information that contradicts their political beliefs, we will select for people that are highly motivated to defend those beliefs.
People who are strongly committed (Brehm & Cohen, 1962; Kiesler, 1971) or have publicly advocated for their position (for a review see Harmon-Jones & Harmon-Jones, 2008), for example, will have a strong driving force that motivates them to defend their political beliefs whereas politically unaffiliated people are relatively less likely to be similarly motivated. In the context of this project, the restraining force is the perceived likelihood of successfully refuting the belief-inconsistent information. Thus, people with low perceived knowledge should be less likely to engage with belief-inconsistent information than people with high perceived knowledge because their lower likelihood of successfully counter-arguing it will act like a restraining force.
Perceived Knowledge and Preference for Belief-Inconsistent Information
It is well known that people tend to prefer information that is consistent with their existing beliefs rather than information that is inconsistent with them (Eagly & Chaiken, 1993;
for a meta-analysis see Hart et al., 2009). Although the research on selective exposure spans many domains (e.g. personality feedback, majority influence, consumer behavior), political information exposure has been a popular topic in this area. Stroud (2008) found that liberals follow news sources that correspond to liberal positions whereas conservatives follow news sources that correspond to conservative positions. Further, Sweeney and Gruber (1984)
demonstrate that Richard Nixon supporters demonstrated less interest and knowledge about the Watergate affair than people who supported a different candidate. People who supported Nixon were motivated to defend their positive view of him and avoiding negative information about him was one way to maintain a favorable attitude towards him.
People demonstrate this preference for belief-consistent information in order to avoid self-threat (Steele, 1988) or psychological discomfort (Festinger, 1957; Brehm & Cohen, 1962;
Blanton, Strauts, & Perez, 2012). For example, Steele (1988) explains that people are motivated to maintain a positive view of the self and therefore belief-inconsistent information can be threatening and unpleasant. If a great math student receives a poor grade on a test in which she expected to do well, this elicits self-threat that she may actually not be as good a math student as she believed. The same idea applies to peoples’ political positions. If political partisans are confronted with strong negative information about their party’s platform, it is threatening because it raises the idea that they are not informed. Thus, she will be motivated to avoid the threat-arousing information.
However, exceptions occur to the general preference for belief-consistent information, particularly when belief-inconsistent information is weak (Lowin, 1967) or if the person is highly
confident that they can defend their views (Albarracín & Mitchell, 2004). For example, Lowin (1967) created and distributed brochures to people during the 1964 presidential campaign that argued in favor of one of the candidates. He manipulated whether the information was consistent or inconsistent with the participant’s political attitudes and ease of refutation (weak vs. strong arguments) to determine whether participants would respond to the offer of more information or ignore it. People accepted the offer of more information arguing in favor of their own party’s candidate more frequently than information favoring the opposing party’s candidate, but only when the arguments were strong. When the arguments were weak, however, people preferred the information arguing in favor of their opposing party’s candidate to information favoring their own party’s candidate. Put simply, people prefer strong information that supports their beliefs or weak information that goes against their beliefs. People seek out information that goes against their beliefs when they can easily refute it. As opposed to strong belief-inconsistent information, weak belief-inconsistent information has a smaller chance of causing cognitive conflict.
Moreover, refuting a weak opposing argument further proves the superiority of one’s own argument. Thus, counter-arguing is another way of protecting oneself from the cognitive conflict resulting from belief threat.
Although people may prefer belief-inconsistent information when it is weak as opposed to strong, Tormala and Petty (2002) demonstrated that participants feel more certain after successfully counter-arguing strong arguments than weak arguments. This is because defeating strong criticisms reifies one’s existing beliefs whereas counter-arguing weak arguments does not.
Similarly, counter-arguing against a high-status source increases attitude certainty but counter- arguing a low-status source does not (Tormala & Petty, 2004). People seem inclined to reason,
“if my position withstands strong attacks, it must be a good position!” This suggests that counter-
arguing high-quality information provides an epistemic function—assuring people that their committed beliefs remain unchallenged.
As the above evidence suggests, successfully counter-arguing against strong, but not weak, information makes people more certain of their own beliefs. Thus, people who believe they have the necessary knowledge should choose to argue against strong belief-inconsistent information because refuting it provides them with an opportunity to increase their certainty. On the other hand, people with low perceived knowledge believe they lack the political knowledge to counter-argue strong information, making weak belief-inconsistent information preferable.
This leads us to expect that people who have high perceived political knowledge, and thus a high driving force, would prefer strong rather than weak belief-inconsistent information whereas people with low political knowledge, and thus a low driving force, would prefer weak rather than strong belief-inconsistent information.
The research reported here is novel and innovative in two ways. First, whereas past research has focused on the consequences of counter-arguing, this research focuses on preferences for counter-arguing. Second, I examined the role of perceived knowledge, which helps us investigate whether people prefer to refute strong belief-inconsistent information if they believe they have the knowledge to do so. This sheds light on factors that influence the defensive biases people prefer, such as whether people avoid or engage with information that contradicts their political beliefs.
Overview of the Current Research
In three interrelated studies, I tested how people defended their political attitudes as a function of their perceived political knowledge. In Study 1, I measured perceived knowledge on four issues and manipulated whether people chose between arguing against a weak versus
moderate argument or a moderate versus strong argument. In Study 2, I measured perceived political knowledge and manipulated the strength of the belief-inconsistent information people received (i.e. weak versus strong). Study 2 is novel because the dependent measure tested whether people chose to engage with or ignore the argument (i.e. complete an unrelated task) as opposed to choosing between two arguments. In Study 3, I administered a political knowledge quiz, assigned participants a fake score, and manipulated the ostensible scores of potential disagreeing discussion partners to see if participants chose to engage in a debate with someone who was slightly versus substantially less (or slightly vs. substantially more) politically
knowledgeable than themselves.
Power Analysis
The novel paradigm of the reported studies made it difficult to determine an informed effect size estimate. According to Cohen (1988), a rule-of-thumb convention for a small odds ratio in the social sciences is 1.49. We used the G*power software to compute the necessary effect size needed to observe this effect with 80% power (for details on G*Power software, see Faul, Erdfelder, Buchner, & Lang, 2009). Assuming an odds ratio of 1.49, we needed a minimum of 214 participants per study.
Chapter 2: Study 1
Study 1 explored people’s preferences for arguing against a strong versus relatively weaker argument. This study sought to test if people choose to argue against stronger arguments to the degree that they are knowledgeable on that issue. In order to test the boundary conditions of preferring to argue against stronger arguments (i.e. preferring to counter stronger arguments up to a point), I manipulated whether participants chose between arguing against a weak versus moderate argument or between a moderate versus strong argument. If perceived issue knowledge
predicts preferences for countering stronger arguments, then this study is an appropriate test of how this relationship changes depending on the relative strength of the available arguments.
Method Participants
Four hundred and fifty-six people (45% female, Mage = 33.45, SDage = 9.65) were recruited through Amazon Mechanical Turk. The TurkPrime software was used to launch the study and restrict access to people in the United States, with a minimal task approval rate of 85%, and a minimum of 100 tasks completed on Mechanical Turk (Litman, Robinson, &
Abberbock, 2016). Six participants were removed for failing the attention check (see Materials section below; also see Oppenheimer, Meyvis, & Davidenko, 2009). The final sample contained 450 participants.1
Procedure
Participants first answered questions about their positions and perceived political knowledge on 4 policy issues (gun control, death penalty, junk food tax, and carbon tax). Next, participants made a series of 4 forced-choices in which they chose between arguing against a relatively stronger versus weaker message criticizing their political views. For each of the four issues, participants were randomly assigned to choose between a weak and moderate argument or a moderate and strong argument. This was repeated for all four issues. Then participants
completed a political knowledge quiz assessing their identification of political public figures as well as their political party knowledge. Lastly, I assessed participants reported behaviors about the frequency of engaging with people who agree or disagree with them on political matters.
Materials
1 Small fluctuations in sample size (2-4 participants) are from analyzing issues separately, in which a given participant had a missing response for one issue but not others. Such differences are reflected in the degrees of freedom reported in each analysis.
Attrition Remedy. After reading a general description of the study, participants completed a task aimed at reducing participant attrition (see Zhou & Fishbach, 2016).
Participants read “Many mTurk workers tend to quit once they see the tasks. If a sizable number of people quit the survey halfway, the data quality of that survey would be
compromised. However, our research depends on good quality data. Thus, before taking this survey, please make sure you do not mind completing the tasks described on the previous page.”
Then participants were instructed, “Please type below ‘I will complete the full survey’ so that we can ensure your data will be of high quality. Thank you!” and then there was a text box for them to type the requested sentence and another that asked for their mTurk identification number.
Attention Check. I used an attention check to ensure that participants were not
mindlessly responding (Goldberg & Carmichael, 2017). Participants were shown a paragraph of text that appeared to be asking participants to select the political activities in which they
regularly engage (e.g. signing petitions, protesting). After three sentences, the message instructed the participant to ignore the instructions and click on an option they would otherwise never choose (“Run for president of the USA”) and ignore all other options.
Issue Attitudes. Participants indicated their position on 4 issues (Gun control, Junk food tax, Carbon tax, and Death penalty). For example, “Do you favor or oppose stricter gun control laws?” (Favor vs. Oppose). For each issue attitude, participants rated their perceived issue knowledge with 4 items (e.g. “How knowledgeable do you feel on this issue?”; “To what extent do you feel you know the most relevant facts about this issue?”; “How much information do you feel you have on this issue?”; “To what extent do you feel you have expertise on this issue?”; 1 = Not at all, 9 = Extremely; α Gun control = .90, α Junk food tax = .94, α Carbon tax = .95, α Death penalty = .94;
see Rios, Goldberg, & Totton, 2017). Participants also rated the strength of their position (1 =
Extremely oppose, 9 = Extremely favor). Finally, participants rated the importance of the issue, their attitude correctness, and their attitude clarity (all scales; 1 = Not at all, 9 = Extremely).
Choice Tasks. Participants were told that they were going make a series of 4 choices between two tasks on each page, and that they would actually complete one of the tasks that they chose. Each choice was between arguing against a relatively stronger and a relatively weaker criticism of the participant’s issue position. For each of the four issues, participants were randomly assigned to read a weak and moderate argument or a moderate and strong argument against their position on that issue (for all arguments, see Appendix A).
The survey software automatically displayed the appropriate stimuli (i.e. belief-
inconsistent arguments) to participants. Participants were asked to choose which argument (weak vs. moderate; moderate vs. strong) they preferred to argue against. They repeated this choice task for each issue until all four issues were completed. After they completed their last choice, they proceeded to the next screen where the argument they had chosen for the death penalty
reappeared along with an essay text box that allowed them to respond to their chosen argument.
The strength of the arguments was determined by pilot testing (N = 75). As in the full study, participants in the pilot were randomly assigned to read a weak and moderate argument or a moderate and strong argument against the participant’s position. Weak arguments were defined by the use of only assertions and anecdotes without empirical evidence. Strong arguments were defined by the use of multiple kinds of strong empirical evidence from authoritative sources in which the evidence points to large impacts in the research findings. Moderate arguments were defined by the use of generic sources (e.g. scientists) that pointed to some impact of the findings cited, but less certain and powerful than those cited in the strong arguments. Participants were asked to rate the strength of each argument they received (1 = Extremely weak, 7 = Extremely
strong). In all cases, a paired samples t-test determined that the moderate argument was rated significantly stronger than the weak argument and the strong argument was rated significantly stronger than the moderate argument (see Tables 1a and 1b).
Political Agreement/Disagreement. Four questions measured people’s general exposure to agreeing and disagreeing others on social media and in person. For example, “How often do you [AGREE/DISAGREE] with political opinions and contents that other people post on social media?” (Never, Only sometimes, Often, Always or nearly always). Participants also answered the same two questions, one asking about agreement and one about disagreement, regarding their opinion exposure in face-to-face interactions. These items were included for exploratory
purposes.
Political Knowledge Quiz. Participants answered 10 factual questions on public figure identification and party knowledge. Questions gauging public figure identification asked about various people in office (e.g. “John Roberts. What job or political office does he now hold?”;
“Michael Pence. What job or political office does he now hold?”). Questions gauging party knowledge asked participants to indicate whether a given position is a Democratic or Republican policy position (e.g. “Which party is generally more supportive of restricting abortion?”; “Which party is generally more supportive of a ‘path to citizenship’?”). We administered this measure after the forced choices task to ensure that taking the quiz will not affect people’s choices about whether to engage with weak, moderate, or strong arguments. Objective knowledge is unlikely to be affected by the choices task and therefore it was preferable to administer the quiz subsequent to the choice tasks.
Demographics. Lastly, participants completed demographic information such as their age, sex, and political ideology (1 = Very liberal, 7 = Very conservative).
Results
I hypothesized that perceived issue knowledge would predict preferences for arguing against the stronger argument when the choice was between a weak and moderate argument and that this effect would disappear when the choice was between a moderate and strong argument.
To test this hypothesis, I entered Condition (0 = Weak vs. Moderate; 1 = Moderate vs. Strong), Issue Perceived Knowledge, and the interaction term into a logistic regression model predicting argument choice (0 = Weaker argument; 1 = Stronger argument). I ran the same model
separately for each issue (Gun control, Junk food tax, Carbon tax, and Death penalty).
For the issue of gun control, the Condition × Perceived Knowledge interaction was not significant, (b = -.09, SE = .11), z(442) = -.82, p = .412, 95% CI [-.31, .13]. However, the main effect of Condition was significant, (b = -.40, SE = .20), z(442) = -2.00, p = .045, 95% CI [-.78, - .01]. That is, when participants had a choice between arguing against a weak or moderate argument, participants chose the weaker argument 57% of the time. When they had a choice between a moderate and strong argument, participants preferred the relatively weaker argument (i.e. moderate argument) even more strongly, 67% of the time, OR = 1.49, 95% CI [1.01, 2.19].
The main effect of Perceived Knowledge was not significant, (b = .03, SE = .06), z(442) = .60, p
= .550, 95% CI [-.08, .14].
For the issue of the junk food tax, the Condition × Perceived Knowledge interaction was not significant, (b = .10, SE = .09), z(444) = -1.09, p = .275, 95% CI [-28, .08]. The main effect for Condition was not significant, (b = .17, SE = .19), z(444) = .89, p = .372, 95% CI [-.21, .56].
The main effect of Perceived Knowledge was marginal such that higher perceived knowledge was associated with higher odds of choosing to argue against the relatively stronger argument, (b
= .08, SE = .05), z(444) = 1.73, p = .084, 95% CI [-.01, .17].
For the issue of the carbon tax, the Condition X Perceived Knowledge was significant, (b
= .22, SE = .09), z(446) = 2.44, p = .015, 95% CI [.04, .40]. When participants reported low perceived knowledge of carbon taxes (-1SD), the difference between conditions was significant, (b = -.85, SE = .28), z(446) = -2.99, p = .003, 95% CI [-1.40, -.29]. That is, when participants received a choice between a moderate and strong argument, it led to 2.34 times the odds of choosing the relatively weaker argument than when participants received a choice between a weak and moderate argument, OR = 2.34, 95% CI [1.34, 4.07]2. When participants reported high perceived knowledge of carbon taxes (+1SD), the difference between conditions disappeared, (b
= .12, SE = .27), z(446) = .43, p = .667, 95% CI [-.41, .65]. See Figure 1 for a depiction of the interaction. The main effect of Condition was marginal, (b = -.37, SE = .19), z(446) = -1.89, p = .059, 95% CI [-.75, .01]. That is, people preferred the relatively weaker argument 53% of the time when their choice was between a weak and moderate argument but their preference for the weaker argument increased to 62% when their choice was between a moderate and strong argument.
For the issue of the death penalty, the Condition × Perceived Knowledge was not
significant, (b = .01, SE = .10), z(446) = -.15, p = .879, 95% CI [-.20, .17]. Neither main effect of Condition nor Perceived Knowledge was significant (ps > .4)
Discussion
Hypotheses for Study 1 were not confirmed. The main theoretical argument received partial confirmation for only one issue, the carbon tax. When participants reported low perceived knowledge on the carbon tax, their preferences for the relatively weaker argument became significantly stronger when they were given a relatively stronger set of argument choices
2 The signs of the odds ratios reported for this interaction were reversed for ease of explanation and interpretation.
(moderate vs. strong) compared to when they were given a weaker set of choices (weak vs.
moderate). This increase in preference for the weaker argument can be interpreted as imposing a greater restraining force on participants, thereby increasing their desire to avoid the stronger argument. Thus, one’s desire to avoid the strong argument should be enhanced when the argument is especially strong (in the moderate vs. strong condition) compared to when it is moderately strong (in the weak vs. moderate condition). When participants reported high perceived knowledge on the carbon tax, however, this preference disappeared and leaned towards (although non-significantly) a preference for relatively stronger arguments. In effect, participants had a higher driving force to defeat the available arguments and were therefore less influenced by the relatively more difficult set of arguments. Although this does not exactly conform to the original hypothesis, it lends some merit to the idea that people with high perceived knowledge are more likely than people with low perceived knowledge to seek out stronger arguments to argue against.
There was a main effect of condition for both gun control and carbon tax. In both cases the main effect demonstrated that people modestly preferred to argue against a weak argument to a moderate argument but preferred to argue against the moderate argument at a substantially higher rate when the alternative was a strong argument. Simply put, people demonstrated stronger preferences for the relatively weaker argument when they chose between a stronger set of arguments (e.g. moderate vs. strong) compared to when they choose between a weaker set of arguments (e.g. weak vs. moderate). The fact that the effects of condition were modest and not moderated by perceived knowledge in three of the four issues suggests that the above findings be interpreted with caution.
A potential methodological flaw that may explain why this study’s hypotheses were not confirmed is that participants’ perceived issue knowledge might not adequately capture whether participants believe they could refute the specific information in the belief-inconsistent
arguments. For example, it is entirely possible that people with high perceived knowledge prefer to argue against stronger arguments they believe they can refute, but our perceived knowledge measure did not assess whether participants believed they could refute the specific message they received in the study. One way to address this issue is to include questions that ask participants about their perceived ability to counter-argue the specific message they receive. If indeed the main flaw of Study 1 was that the perceived knowledge measure did not appropriately measure one’s perceived ability to counter-argue to messages they read, then using a better measure in Study 2 would produce more convincing results. For example, it may be better to ask, “how likely are you to effectively counter-argue the argument above?” than the general knowledge questions we asked in Study 1 such as “how much knowledge do you feel you have on this issue?” because the former asks about the specific argument in the task whereas the latter asks about the participants’ general perceived knowledge.
Additionally, more research is needed to explore the factors that predict whether someone will use one defense strategy over another strategy (e.g. counter-argue versus ignore) as opposed to arguing against one message over another. Giving participants more flexibility in how to defend their beliefs should give us more insight into what they would do if they were unconstrained by the experimental procedures. This is the main contribution of Study 2.
Chapter 3: Study 2
The purpose of Study 2 was to test how the strength of a belief-inconsistent message influences how people defend against it, such as deciding whether to counter-argue the message
or ignore it. Study 2 explored a situation in which participants rated their perceived issue knowledge, were randomly assigned to receive either a weak or strong argument against their issue position, gauged their ability to counter-argue the argument, and had the choice between counter-arguing the message on the screen or to complete an unrelated task. If people are more interested in counter-arguing against stronger arguments if they have the knowledge to do so, then I hypothesize that perceived knowledge will predict preferences for counter-arguing when participants receive a strong belief-inconsistent message but will predict preferences for ignoring the message when it is weak.
Method Participants
Four hundred and twenty-eight people participated in this study through Amazon’s Mechanical Turk. The TurkPrime software (Litman, Robinson, & Abberbock, 2016) was used to make the same participant restrictions as Study 1 (from the U.S.A., minimum 85% task approval, minimum 100 tasks completed). Ten participants were removed for failing the attention check.
The final sample contained 418 participants3. Procedure
In Study 2, participants answered the same questions as in Study 1 that tap into their policy positions and their perceived issue knowledge for the issues of gun control and the death penalty. Next, participants were randomly assigned to receive a strong or weak argument against their political position on one of the issues. Participants then rated their perceived likelihood of successfully counter-arguing the message and then chose whether they preferred to provide a
3 Small deviations from this sample size for some analyses are due to participant missing data from some variables but not others. I sought to maximize the sample size by using all available data for that particular analysis.
written argument against the message or complete a writing task unrelated to their political beliefs. The dependent variable was participants’ decision to argue versus ignore the message.
Materials
Argument Manipulation. Participants were randomly assigned to receive either a weak or strong argument that criticized their position on either gun control or the death penalty. We reduced this study to two issues because gun control and the death penalty are likely to be more familiar to participants, and therefore participants are more likely to have a clearer position on these issues than on the issues of junk food or carbon taxes. Additionally, given the same sample size, it is more statistically powerful to use two issues instead of four. The strong and weak arguments for these issues were identical to those in Study 1.
Choice Task. Participants were instructed to choose between arguing against the
information or completing another unrelated task that has been equated for interest and difficulty.
Participants read, “The task on this page is to spend the next 3 minutes arguing against the message below. However, we understand that some participants might want to complete a different task instead. We pretested several other tasks to ensure they take the same amount of time and effort so you may choose to complete the current task or an alternative unrelated task.”
Expectations of Effective Counter-Arguing. After reading the argument criticizing their position on the randomly selected issue (Gun control or the Death penalty), participants answered four questions measuring their expectations of effective counter-arguing (see Albarracín & Mitchell, 2004). Participants rated how likely they were to “effectively counter- argue the argument above,” “effectively refute the argument above”, “identify problems in the argument above”, “know counter-arguments to the argument above” (1 = Extremely unlikely, 9 =
Extremely likely). These four items were averaged to form a composite measure of expectations
of effective counter-arguing (α = .95)
Instructions for Chosen Task. If participants chose to complete the counter-arguing task, they were instructed on the next page “Please spend the next three minutes arguing against the message above.” If participants chose to complete the unrelated task, they were instructed on the next page “Please spend the next three minutes describing what political issue is most
important to you and why.” After making their choice, participants completed their chosen task and answered basic demographic questions.
Results
I hypothesized that, when participants receive a strong criticism of their position,
perceived issue knowledge would positively predict a preference for arguing against the criticism (vs. choosing to do the unrelated task). When participants receive a weak criticism, I predicted that perceived issue knowledge would negatively predict a preference for arguing against the criticism (i.e. preference for doing an unrelated task). To test this hypothesis, I first tested the Condition (Weak vs. Strong) × Perceived Knowledge × Issue (Gun control vs. Death penalty) interaction to ensure that the hypothesized effects were not moderated by issue. It was not significant, (b = -.19, SE = .28), z(410) = -.68, p = .499, 95% CI [-.74, .36] and therefore I collapsed across issue and tested the Condition (Weak vs. Strong) × Perceived Knowledge interaction predicting task choice (0 = Counter-argue task, 1 = Unrelated task). Although the interaction was not significant, (b = -.19, SE = .14), z(414) = -1.39, p = .163, 95% CI [-.46, .08], the pattern of results lend partial support to my theoretical argument. Unexpectedly, when participants received a weak argument, perceived knowledge significantly predicted preferences to counter-argue over completing an unrelated task, (b = -.19, SE = .08), z(414) = -2.47, p = .014,
95% CI [-.34, .04]. That is, each one unit increase in perceived knowledge led to 21% greater odds of choosing to counter-argue, OR = 1.21, 95% CI [1.04, 1.41]. In line with the hypothesis, however, when participants received a strong argument, this relationship got even stronger, (b = - .38, SE = .11), z(414) = -3.35, p < .001, 95% CI [-.61, -.16]. That is, each one unit increase in perceived knowledge led to a 47% greater odds of choosing to counter-argue, OR = 1.47, 95%
CI [1.17, 1.84] (See Figure 2).
The main effect of condition was significant such that participants were significantly more likely to choose the unrelated task (i.e. ignore the argument) when they received a strong argument than when they received a weak argument, (b = 1.03, SE = .25), z(414) = 4.14, p <
.001, 95% CI [.54, 1.52]. That is, when participants received a strong argument it led to 2.80 times the odds of choosing the unrelated task over counter-arguing, OR = 2.80, 95% CI [1.72, 4.57]. The main effect of perceived knowledge was significant such that participants were significantly more likely choose to counter-argue as perceived knowledge increased, (b = -.28, SE = .07), z(414) = -4.16, p < .001, 95% CI [-.41, -.15].
In an attempt to more precisely test if people seek to counter-argue messages only to the degree they believe they could do so successfully, I used the PROCESS macro in SPSS (Model 4; Hayes, 2013) to test if the relationship between Condition (X; Weak vs. Strong) and Task Choice (Y; 0 = Counter-argue task, 1 = Unrelated task) was mediated by their perceived expectations of effective counter-arguing (M; continuous). Compared to when participants received a weak argument, giving participants a strong argument increased participants’
preferences for the unrelated task indirectly through their perceptions that they could not effectively counter-argue the message. Put another way, giving participants a weak belief- inconsistent message increased their belief that they could effectively counter-argue it, which in
turn predicted preferences for counter-arguing over completing the unrelated task (i.e. ignoring the message; see Figure 3). Bootstrapping revealed that the indirect effect was significantly different from zero (b = .61, SEboot = .14), 95% CIboot [.38, .93].
Discussion
Study 2 findings confirm the hypothesis that, when participants received a strong argument, perceived knowledge predicted preferences for counter-arguing as opposed to completing an unrelated task. I did not expect to find that, when the argument was weak, perceived knowledge also predicted preferences for counter-arguing. Although this was not expected—and the interaction not quite significant—these results provide some support for my theoretical argument that people prefer to argue against a strong argument to the degree they believe they have the knowledge to do so and that this relationship was somewhat weaker when participants received a weak argument. I proposed that this is because people become more certain after arguing against strong (as opposed to weak) arguments (Tormala & Petty, 2002;
2004). Thus, people should see greater value in counter-arguing a strong criticism of their beliefs than a weak one and that people would choose to counter-argue accordingly. It is also possible, however, that perceived knowledge was a weaker predictor in the weak condition because most people believed they could successfully counter-argue.
It is worth noting that there was no level of perceived knowledge at which participants’
majority choice was to counter-argue. In other words, participants always chose to ignore at a higher rate than counter-argue. However, as perceived knowledge increased, participants’
preferences to counter-argue also increased.
The analysis of mechanism was promising. Compared to when participants received a strong argument, receiving a weak argument cause participants to increase their expectations that
they could effectively counter-argue which in turn led participants to increase their preference for counter-arguing instead of ignoring. The fact that expectations of effective counter-arguing strongly predicted preferences for counter-arguing suggests that there is epistemic value in counter-arguing over ignoring. If people’s cognitive goal is to feel certain about their beliefs, Cognitive Energetics Theory would predict that people will invest the necessary energy to maintain certainty even if the task is more laborious than alternative tasks. Ignoring the
information leaves the cognitive discrepancy intact whereas counter-arguing does not. Thus, it is more defensively advantageous to counter-argue than to ignore if one believes they can counter- argue successfully. In terms of CET, people who believe they will be successful in counter- arguing have a high driving force and will be more likely to choose the task that fits their cognitive goal of maintaining attitude certainty.
Taken together, these findings are encouraging for future research that will help identify the factors that predict people’s choice of belief-defense. Given that people still largely chose to ignore the belief-inconsistent information suggests that information avoidance will continue to be an obstacle in fostering intergroup communication.
One limitation of Study 2 is that participants got to see the argument that they might choose to argue against (versus ignore) before making their decision. Thus, they were likely fully aware of whether or not they could successfully counter-argue the message. When engaging with disagreeing others, it is not always clear what the other person might argue in defending their position, so this might breed more or less perceived “risk” of arguing with the other person. In Study 3, I use a novel paradigm to manipulate the perceived issue knowledge of the participant’s potential debate partner, which should provide more insight into the differences between
choosing arguments versus choosing someone to debate.
Chapter 4: Study 3
In Study 3 I investigated people’s choice of debate partner depending on the partner’s ostensible issue knowledge. In Studies 1 and 2 I manipulated participants’ exposure to arguments of varying strength. In Study 3, on the other hand, I manipulated participants’ choices of a debate partner. I reasoned that people would be less interested in debating someone with substantially lower knowledge than themselves because there would be little to epistemically gain from the debate. If people have a high driving force to maintain or even increase attitude certainty, then this goal would be best achieved by arguing against someone who is relatively high in
knowledge (Tormala & Petty, 2002; 2004), but not so high as to pose a risk of decreasing certainty. Thus, I expected that participants would choose a debate partner with similar issue knowledge over one with substantially lower knowledge. However, I expected that when people had a choice between someone with similar issue knowledge or substantially more knowledge, it would create a greater risk of their beliefs being truly challenged if they were to choose the substantially more knowledgeable partner. Therefore, people would again prefer a debate partner with similar knowledge to oneself.
Method Participants
Four hundred and fifty two people participated in this study through Amazon Mechanical Turk (46% female, Mage = 36.22, SDage = 11.94). Like Studies 1 and 2, I used the TurkPrime software (Litman, Robinson, & Abberbock, 2016) to limit study access to people in United States, with a minimum task approval of 85%, and a minimum of 100 tasks completed on Mechanical Turk. Twelve people were removed from analyses because they did not believe the discussion would be real.
Procedure
First, participants answered basic demographic questions such as age and sex. Then participants indicated their position on gun control and the death penalty. For each issue, participants answered questions measuring their perceived knowledge for each issue. Then participants took a quiz that ostensibly measured their knowledge on the death penalty.
Participants then proceeded to the next page that stated that their results were being calculated.
After being given false feedback about their knowledge score (which was identical for all participants), I manipulated participants’ choice between two debate partners, both of whom always had the opposite opinion of the participant. The two potential debate partners’ death penalty knowledge scores were also displayed. Participants were randomly assigned to choose between 1) a discussion partner with a similar death penalty knowledge score and one with a significantly lower score or between 2) a discussion partner with a similar death penalty knowledge score and one with a significantly higher score. Participants made their choice, completed a suspicion probe, and then were informed that there would be no debate.
Materials
Demographics. First, participants indicated their age, sex, and political identification (1
= Very liberal, 7 = Very conservative).
Issue Positions. Then participants answered the same questions from Studies 1 and 2 on their position on gun control and the death penalty (Favor vs. Oppose) and four items for each issue measuring their perceived knowledge (α = .93).
Political Knowledge Quiz. Participants took a political knowledge multiple choice quiz and were given a score that ostensibly measured their political knowledge on the death penalty.
The quiz was constructed using easy, moderate, and difficult questions so that assigning all
participants a moderate score (i.e. 67th percentile) was believable (for all quiz questions and answer choices, see Appendix B). For example, an easy question was “What is the most accurate definition of capital punishment?” whereas a difficult question was “What is the average cost to tax payers of life in prison?” On the following page they read “Your quiz results are being calculated. Please move on to the next page to view your results.”
Once participants move on to the next page, they read “The results of this quiz indicate that you answered 6/10 of the questions correctly. This means that you scored higher than 67%
of your peers.” The participant’s score was held constant for all participants.
Debate Partner Manipulation. On the following page, participants read that they would be debating the topic of the death penalty with another worker from Mechanical Turk (for a similar procedure, see Rios, Goldberg, & Totton, 2017). They read:
“In this part of the study, we are interested in how people talk about controversial issues with others. You will participate in an online discussion with another person who is currently logged into this study.
Recently, we have taken advantage of a special new feature in our survey design system, which allows us to connect you with people who are also currently participating in this study. Because this is such a massive study, and online experiments on Mechanical Turk are so convenient and popular, it generally takes less than a minute to connect you via our chat feature to another participant who has the opposite opinion as you do on the death penalty.
For this discussion, you can choose between two people that have been randomly selected by our survey system. Please go to the next screen to make this choice.”
On the following page, participants were reminded of their knowledge score and told that a choice of two possible debate partners were randomly selected from which they can choose to debate. For both potential debate partners, an 8-digit identification number, the person’s position on the death penalty, and the person’s quiz score (measuring knowledge of the death penalty) were displayed. The debate partners’ scores were shown in a similar format to the participant’s
score. For example, “Quiz score: 3/10, which is better than 36% of your peers.” Participants were randomly assigned to one of two sets of partner choices. The choice between discussion partners was either 1) a discussion partner with a similar knowledge score (64th percentile) vs.
one with a substantially lower score (36th percentile) or 2) a discussion partner with a similar knowledge score (70th percentile) vs. a substantially higher score (98th percentile). In all cases, participants chose between two discussion partners that both have the opposite opinion on the issue that will be discussed.
Suspicion Probe. After participants made their choice of discussion partner, they were given the following suspicion probe: “Before we begin, please jot down as specifically as you can what you think will happen during the online discussion.” If participants expressed that they did not believe the discussion would be real, they were removed from the sample (see
Participants section above). Then participants were told that they would not be having a discussion with another person and were informed about the purpose of the study.
Results
I hypothesized that participants would choose a discussion partner with a similar knowledge score (i.e. close to 67th percentile) when the alternative choice was a partner with substantially lower knowledge. Further, I hypothesized that participants would choose a
discussion partner with a similar knowledge score when the alternative choice was a partner with substantially higher knowledge. To test this hypothesis, I entered condition (0 = extremely low knowledge alternative; 1 = extremely high knowledge alternative) into a binary logistic
regression predicting choice of partner (0 = similar to own knowledge; 1 = alternative). Further, I added participants’ ratings of their perceived knowledge on the death penalty as a covariate to control for individual differences in perceived knowledge that existed before the fake quiz
feedback was given to participants. As expected, participants chose the discussion partner with similar knowledge on the death penalty more frequently than a substantially less knowledgeable alternative. Unexpectedly, however, participants chose the substantially more knowledgeable alternative partner more frequently than the similar knowledge partner. Put simply, participants in both conditions chose the more knowledgeable partner. The main effect was significant, (b = .15, SE = .05), z(437) = 3.11, p = .002, 95% CI [.05, .24]. Going from the condition in which the alternative partner was substantially less knowledgeable to the condition in which the alternative partner was substantially more knowledgeable led to 16% greater odds of choosing the
alternative, OR = 1.16, 95% CI [1.06, 1.27]. That is, people preferred a similarly knowledgeable partner to an alternative that was substantially less knowledgeable but increased their preference for an alternative partner when they were substantially more knowledgeable than the participant.
I also tested the Condition × Perceived Knowledge interaction to test if the effect of condition was weaker or non-existent for participants with low perceived knowledge and
stronger for participants with high perceived knowledge. I ran the same binary logistic regression model with the Condition × Perceived Knowledge interaction term. The interaction was
marginally significant, (b = .16, SE = .10), z(436) = 1.62, p = .105, 95% CI [-.03, .35]. The effect of condition was not significant when participants had low perceived knowledge (-1SD;
perceived knowledge score = 3.67 out of 9), (b = .28, SE = .27), z(436) = 1.04, p = .300, 95% CI [-.25, .82]. However, the effect of condition got progressively stronger as perceived knowledge went up and became significant once participants reached 4.61 out of 9 on the perceived knowledge scale, (b = .43, SE = .22), z(436) = 1.96, p = .050, 95% CI [.00, .86]. This effect remained significant and continued to get stronger for participants as they reported higher
perceived knowledge (i.e. +1SD; perceived knowledge score = 7.73 out of 9), (b = .92, SE = .28),
z(436) = 3.29, p = .001, 95% CI [.37, 1.46]. Put simply, participants in both conditions chose the more knowledgeable partner but this effect was negligible when participants had low perceived knowledge and was strong when participants had high perceived knowledge (see Figure 4).
Discussion
The Study 3 findings show strong support for the idea that people are uninterested in arguing with someone who has substantially lower knowledge than they do. Perhaps this is because there is little to epistemically gain from a conversation with someone who is very low in issue knowledge—especially given that participants, on average, were above the midpoint on their perceived knowledge (M = 5.69 out of 9). Previous research shows that, when people counter-argue against weak arguments or against low-status sources, they do not become any more certain of their beliefs (Tormala & Petty, 2002; 2004). I expected that participant
preferences for a relatively more knowledgeable partner would be diminished when the partner had a substantially higher knowledge score (as opposed to one that is similar to the participant).
Contrary to my expectations, however, participants preferred a partner with a substantially higher knowledge score than one that had a similar knowledge score to the participant. My primary argument was that people prefer challenging arguments to weak ones only to the degree that they believe they can successfully counter-argue the opposition. This main effect suggests that this preference extends to arguing with others who are ostensibly more knowledgeable on the topic.
This is bolstered by previous research that shows that people become more certain after counter- arguing against strong arguments or against high-status sources (Tormala & Petty, 2002; 2004).
It is possible that participants in Study 3 had anticipated how certain they might feel after arguing with a more versus less knowledgeable person.
The moderating effects of participant’s perceived knowledge provides clear support for my primary argument that people prefer to argue against knowledgeable others but only if they believe they are knowledgeable enough to successfully do so. The main effect of condition discussed above shows that people prefer to argue against knowledgeable others, but the
interaction with participant perceived knowledge shows that this effect is non-existent for people low in perceived knowledge and is very strong for people high in perceived knowledge. In terms of Cognitive Energetics Theory, people with high perceived knowledge had a sufficiently high driving force that made them overcome the restraining force of arguing against a relatively more knowledgeable discussion partner. People with low perceived knowledge, however, were
constrained by their available knowledge, making them less interested in arguing against a relatively more knowledgeable discussion partner.
Taken together, these findings corroborate and extend research that shows that people who believe they can effectively defend their beliefs are more likely to engage with belief-
inconsistent information than people who doubt their ability to defend their beliefs (Albarracín &
Mitchell, 2004). The current research builds on this by demonstrating that people seek out more challenging arguments (i.e. by choosing more knowledgeable partners) even if their discussion partner is ostensibly more knowledgeable than the participant. It would be fruitful to explore the mechanism that drives this effect. Previous research would suggest that people who are high in perceived knowledge might engage in an argument with a disagreeing other with expectations that they will leave the conversation feeling even more confident in their beliefs (Albarracín &
Mitchell, 2004).
Chapter 5: General Discussion
The research reported in this dissertation provides evidence for the idea that people prefer to counter-argue more difficult messages only if they are high in perceived knowledge. In Study 1, participants who believed they knew little about the carbon tax increased their preference for arguing against the weaker argument when they chose between a moderate and strong argument compared to when they chose between a weak and moderate argument. On the other hand, when participants believed that they knew a lot about the carbon tax, there were no differences in argument preference. Although participants overall preferred to argue against the weaker argument, perceived knowledge of the carbon tax emboldened people to take on stronger arguments. It is important to state, however, that this effect be interpreted with caution as it was the only significant interaction out of four issues.
Study 2 demonstrated support for the idea that perceived knowledge predicts desire to counter-argue more strongly when the argument was strong than when it was weak. This
supports the proposition that people who believe they can beat strong arguments will prefer them to weak arguments because they offer a better opportunity to reaffirm one’s sense of certainty (Tormala & Petty, 2002; 2004). Here, for the first time, I demonstrate that people can become more interested in attempting to counter stronger threats to their beliefs (as opposed to weak threats) so as long as they believe they can defeat them. This is consistent with research showing that people with high confidence that they can defend their attitudes will show smaller
information-selection biases in favor of information that supports their beliefs (Albarracín &
Mitchell, 2004).
In order to improve upon Study 1, I included in Study 2 a more specific measure that gauges one’s perceived ability to counter-argue. The analysis of mechanism demonstrated that people’s expectations of effective counter-arguing explain the relationship between argument
